[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"sanity-iu16u4U3yi4VD6sbTiW1ChUYsopW_Y8Xu3NSX1MSxgg":3},{"data":4,"sourceMap":-1},{"info":5,"latest":22},{"_createdAt":6,"_id":7,"_rev":8,"_type":9,"_updatedAt":6,"avatar":10,"bio":15,"employee":16,"name":17,"role":18,"slug":19},"2024-04-04T14:34:48Z","12169ed9-78d5-4f82-adb4-e956348db389","l85YFb2NNiT2L5SEvOgP0n","blogAuthor",{"_type":11,"asset":12},"image",{"_ref":13,"_type":14},"image-7c75ba835850e53201ea979015e8b8f449e126da-400x400-png","reference","I’m currently the Director of AI at Rebuy, a personalized search and recommendations platform for D2C e-commerce brands. Prior to Rebuy, I was a Research Scientist at Alegion. Additionally, I worked for Salesforce Commerce Cloud for two years. ","none","Cameron R. Wolfe, PhD","Director of AI at Rebuy",{"_type":20,"current":21},"slug","cameron-r-wolfe-phd",[23,65,90,115,139,164,192,226,251],{"_id":24,"author":25,"commentCount":30,"comments":31,"excerpt":32,"featureTag":33,"image":34,"publishedAt":38,"slug":39,"sponsored":33,"tags":41,"title":64},"a2ebfc26-d36a-4960-9d55-daef5ff386c4",[26],{"_id":7,"avatar":27,"name":17,"role":18,"slug":29},{"_type":11,"asset":28},{"_ref":13,"_type":14},{"_type":20,"current":21},27,true,"Self-supervised learning is a key advancement that revolutionized natural language processing and generative AI. Here’s how it works and two examples of how it is used to train language models.",null,{"_type":11,"asset":35,"attribution":37},{"_ref":36,"_type":14},"image-c86e442c47972984ddb7781488ad94ee1f0c4c04-12000x6293-jpg","Alexandra Francis","2025-04-28T16:00:00.000Z",{"_type":20,"current":40},"how-self-supervised-language-revolutionized-natural-language-processing-and-gen-ai",[42,49,53,57],{"_createdAt":43,"_id":44,"_rev":45,"_type":46,"_updatedAt":43,"slug":47,"title":48},"2023-05-23T16:43:21Z","wp-tagcat-se-tech","9HpbCsT2tq0xwozQfkc4ih","blogTag",{"current":48},"se-tech",{"_createdAt":43,"_id":50,"_rev":45,"_type":46,"_updatedAt":43,"slug":51,"title":52},"wp-tagcat-se-stackoverflow",{"current":52},"se-stackoverflow",{"_createdAt":43,"_id":54,"_rev":45,"_type":46,"_updatedAt":43,"slug":55,"title":56},"wp-tagcat-llm",{"current":56},"llm",{"_createdAt":43,"_id":58,"_rev":59,"_type":46,"_updatedAt":60,"slug":61,"title":63},"wp-tagcat-ai","fpDTFQqIDjNJIbHDKPBGpV","2025-01-30T16:19:01Z",{"current":62},"ai","AI","How self-supervised learning revolutionized natural language processing and gen AI",{"_id":66,"author":67,"commentCount":72,"comments":31,"excerpt":73,"featureTag":33,"image":74,"publishedAt":77,"slug":78,"sponsored":33,"tags":80,"title":89},"ec74f3ce-0e51-4de2-9492-e6ee736801ed",[68],{"_id":7,"avatar":69,"name":17,"role":18,"slug":71},{"_type":11,"asset":70},{"_ref":13,"_type":14},{"_type":20,"current":21},0,"Want to train a specialized LLM on your own data? The easiest way to do this is with low rank adaptation (LoRA), but many variants of LoRA exist. ",{"_type":11,"asset":75,"attribution":37},{"_ref":76,"_type":14},"image-f4473993106f8912780f3034210971f5dd8391e1-12000x6293-jpg","2025-02-26T17:00:00.000Z",{"_type":20,"current":79},"variants-of-lora",[81,83,85,87],{"_createdAt":43,"_id":44,"_rev":45,"_type":46,"_updatedAt":43,"slug":82,"title":48},{"current":48},{"_createdAt":43,"_id":50,"_rev":45,"_type":46,"_updatedAt":43,"slug":84,"title":52},{"current":52},{"_createdAt":43,"_id":54,"_rev":45,"_type":46,"_updatedAt":43,"slug":86,"title":56},{"current":56},{"_createdAt":43,"_id":58,"_rev":59,"_type":46,"_updatedAt":60,"slug":88,"title":63},{"current":62},"Variants of LoRA",{"_id":91,"author":92,"commentCount":97,"comments":31,"excerpt":98,"featureTag":33,"image":99,"publishedAt":102,"slug":103,"sponsored":33,"tags":105,"title":114},"8dcc8bcc-2220-4803-b6a3-2b9aaf72e598",[93],{"_id":7,"avatar":94,"name":17,"role":18,"slug":96},{"_type":11,"asset":95},{"_ref":13,"_type":14},{"_type":20,"current":21},2,"Wondering how to go about creating an LLM that understands your custom data? Start here. ",{"_type":11,"asset":100,"attribution":37},{"_ref":101,"_type":14},"image-1842af09277124c00aa4c015f2852dba5b225e1a-3000x1575-jpg","2024-12-05T17:00:00.000Z",{"_type":20,"current":104},"four-approaches-to-creating-a-specialized-llm",[106,108,110,112],{"_createdAt":43,"_id":44,"_rev":45,"_type":46,"_updatedAt":43,"slug":107,"title":48},{"current":48},{"_createdAt":43,"_id":50,"_rev":45,"_type":46,"_updatedAt":43,"slug":109,"title":52},{"current":52},{"_createdAt":43,"_id":54,"_rev":45,"_type":46,"_updatedAt":43,"slug":111,"title":56},{"current":56},{"_createdAt":43,"_id":58,"_rev":59,"_type":46,"_updatedAt":60,"slug":113,"title":63},{"current":62},"Four approaches to creating a specialized LLM",{"_id":116,"author":117,"commentCount":72,"comments":31,"excerpt":122,"featureTag":33,"image":123,"publishedAt":126,"slug":127,"sponsored":33,"tags":129,"title":138},"14a8e708-41c6-4757-a2ff-9802f42c99f4",[118],{"_id":7,"avatar":119,"name":17,"role":18,"slug":121},{"_type":11,"asset":120},{"_ref":13,"_type":14},{"_type":20,"current":21},"Here's a (brief) summary of language model finetuning, the various approaches that exist, their purposes, and what we know about how they work.",{"_type":11,"asset":124,"attribution":37},{"_ref":125,"_type":14},"image-99abe272d164c394aa1d28206288f1b0cdac880e-6000x3150-jpg","2024-10-31T14:00:00.000Z",{"_type":20,"current":128},"a-brief-summary-of-language-model-finetuning",[130,132,134,136],{"_createdAt":43,"_id":44,"_rev":45,"_type":46,"_updatedAt":43,"slug":131,"title":48},{"current":48},{"_createdAt":43,"_id":50,"_rev":45,"_type":46,"_updatedAt":43,"slug":133,"title":52},{"current":52},{"_createdAt":43,"_id":54,"_rev":45,"_type":46,"_updatedAt":43,"slug":135,"title":56},{"current":56},{"_createdAt":43,"_id":58,"_rev":59,"_type":46,"_updatedAt":60,"slug":137,"title":63},{"current":62},"A brief summary of language model finetuning",{"_id":140,"author":141,"commentCount":146,"comments":31,"excerpt":147,"featureTag":33,"image":148,"publishedAt":151,"slug":152,"sponsored":33,"tags":154,"title":163},"484c1547-c0ce-40f6-a02b-8e59ec0d4d29",[142],{"_id":7,"avatar":143,"name":17,"role":18,"slug":145},{"_type":11,"asset":144},{"_ref":13,"_type":14},{"_type":20,"current":21},1,"Masked self-attention is the key building block that allows LLMs to learn rich relationships and patterns between the words of a sentence. Let’s build it together from scratch.",{"_type":11,"asset":149,"attribution":37},{"_ref":150,"_type":14},"image-e278dfb29b78d1ba33dc27e79d5906eb53acf972-12000x6300-jpg","2024-09-26T16:00:00.000Z",{"_type":20,"current":153},"masked-self-attention-how-llms-learn-relationships-between-tokens",[155,157,159,161],{"_createdAt":43,"_id":44,"_rev":45,"_type":46,"_updatedAt":43,"slug":156,"title":48},{"current":48},{"_createdAt":43,"_id":50,"_rev":45,"_type":46,"_updatedAt":43,"slug":158,"title":52},{"current":52},{"_createdAt":43,"_id":54,"_rev":45,"_type":46,"_updatedAt":43,"slug":160,"title":56},{"current":56},{"_createdAt":43,"_id":58,"_rev":59,"_type":46,"_updatedAt":60,"slug":162,"title":63},{"current":62},"Masked self-attention: How LLMs learn relationships between tokens",{"_id":165,"author":166,"commentCount":72,"comments":31,"excerpt":171,"featureTag":33,"image":172,"publishedAt":175,"slug":176,"sponsored":33,"tags":178,"title":191},"3e42d4a4-c110-4467-97b1-fc2e8578baa6",[167],{"_id":7,"avatar":168,"name":17,"role":18,"slug":170},{"_type":11,"asset":169},{"_ref":13,"_type":14},{"_type":20,"current":21},"The decoder-only transformer architecture is one of the most fundamental ideas in AI research.",{"_type":11,"asset":173,"attribution":37},{"_ref":174,"_type":14},"image-a88b67b632d885d9634ccc284b8924dd01f9b4ef-6000x3150-jpg","2024-08-22T16:25:00.000Z",{"_type":20,"current":177},"llms-evolve-quickly-their-underlying-architecture-not-so-much",[179,181,183,185,189],{"_createdAt":43,"_id":44,"_rev":45,"_type":46,"_updatedAt":43,"slug":180,"title":48},{"current":48},{"_createdAt":43,"_id":50,"_rev":45,"_type":46,"_updatedAt":43,"slug":182,"title":52},{"current":52},{"_createdAt":43,"_id":54,"_rev":45,"_type":46,"_updatedAt":43,"slug":184,"title":56},{"current":56},{"_createdAt":43,"_id":186,"_rev":45,"_type":46,"_updatedAt":43,"slug":187,"title":188},"wp-tagcat-architecture",{"current":188},"architecture",{"_createdAt":43,"_id":58,"_rev":59,"_type":46,"_updatedAt":60,"slug":190,"title":63},{"current":62},"LLMs evolve quickly. Their underlying architecture, not so much.",{"_id":193,"author":194,"commentCount":72,"comments":31,"excerpt":199,"featureTag":33,"image":200,"publishedAt":203,"slug":204,"sponsored":33,"tags":206,"title":225},"14d5f250-b990-4fc9-ac5f-fe5c38fe8c28",[195],{"_id":7,"avatar":196,"name":17,"role":18,"slug":198},{"_type":11,"asset":197},{"_ref":13,"_type":14},{"_type":20,"current":21},"Retrieval-augmented generation (RAG) is one of the best (and easiest) ways to specialize an LLM over your own data, but successfully applying RAG in practice involves more than just stitching together pretrained models.",{"_type":11,"asset":201,"attribution":37},{"_ref":202,"_type":14},"image-a30b19f63157ddf6ea053fc4c83247b7dcd7152d-13396x7384-jpg","2024-08-15T16:00:00.000Z",{"_type":20,"current":205},"practical-tips-for-retrieval-augmented-generation-rag",[207,209,211,218,220],{"_createdAt":43,"_id":44,"_rev":45,"_type":46,"_updatedAt":43,"slug":208,"title":48},{"current":48},{"_createdAt":43,"_id":50,"_rev":45,"_type":46,"_updatedAt":43,"slug":210,"title":52},{"current":52},{"_createdAt":212,"_id":213,"_rev":214,"_type":46,"_updatedAt":212,"slug":215,"title":217},"2023-10-18T14:37:04Z","df698ce8-b4c2-4d8f-8488-1406a8b7e8e9","Qk9b1CKg885xXVOHAyC7bl",{"_type":20,"current":216},"retrieval-augmented-generation","retrieval augmented generation",{"_createdAt":43,"_id":54,"_rev":45,"_type":46,"_updatedAt":43,"slug":219,"title":56},{"current":56},{"_createdAt":43,"_id":221,"_rev":45,"_type":46,"_updatedAt":43,"slug":222,"title":224},"wp-tagcat-generative-ai",{"current":223},"generative-ai","generative AI","Practical tips for retrieval-augmented generation (RAG)",{"_id":227,"author":228,"commentCount":233,"comments":31,"excerpt":234,"featureTag":33,"image":235,"publishedAt":238,"slug":239,"sponsored":33,"tags":241,"title":250},"8297b220-990a-43cb-9361-8ed67b5291e1",[229],{"_id":7,"avatar":230,"name":17,"role":18,"slug":232},{"_type":11,"asset":231},{"_ref":13,"_type":14},{"_type":20,"current":21},19,"Here’s a simple, three-part framework that explains generative language models.",{"_type":11,"asset":236,"attribution":37},{"_ref":237,"_type":14},"image-bf522896b78c1eebec939ce8c09865c8cfd36882-12000x6300-jpg","2024-06-27T16:20:00.000Z",{"_type":20,"current":240},"explaining-generative-language-models-to-almost-anyone",[242,244,246,248],{"_createdAt":43,"_id":50,"_rev":45,"_type":46,"_updatedAt":43,"slug":243,"title":52},{"current":52},{"_createdAt":43,"_id":44,"_rev":45,"_type":46,"_updatedAt":43,"slug":245,"title":48},{"current":48},{"_createdAt":43,"_id":221,"_rev":45,"_type":46,"_updatedAt":43,"slug":247,"title":224},{"current":223},{"_createdAt":43,"_id":54,"_rev":45,"_type":46,"_updatedAt":43,"slug":249,"title":56},{"current":56},"Explaining generative language models to (almost) anyone",{"_id":252,"author":253,"commentCount":72,"comments":31,"excerpt":258,"featureTag":33,"image":259,"publishedAt":262,"slug":263,"sponsored":33,"tags":265,"title":281},"bf6cc061-dcb9-41cd-bfa4-6d15b2421dd9",[254],{"_id":7,"avatar":255,"name":17,"role":18,"slug":257},{"_type":11,"asset":256},{"_ref":13,"_type":14},{"_type":20,"current":21},"This new LLM technique has started improving the results of models without additional training. ",{"_type":11,"asset":260,"attribution":37},{"_ref":261,"_type":14},"image-806103733e75354ebebd71aa8baf817451aec6e1-1280x672-jpg","2024-04-04T14:31:11.098Z",{"_type":20,"current":264},"how-do-mixture-of-experts-layers-affect-transformer-models",[266,268,270,277,279],{"_createdAt":43,"_id":221,"_rev":45,"_type":46,"_updatedAt":43,"slug":267,"title":224},{"current":223},{"_createdAt":43,"_id":54,"_rev":45,"_type":46,"_updatedAt":43,"slug":269,"title":56},{"current":56},{"_createdAt":271,"_id":272,"_rev":273,"_type":46,"_updatedAt":271,"slug":274,"title":276},"2024-04-04T14:35:24Z","d0a9b59b-c92e-48f3-84ea-4a9d1093b4e3","l85YFb2NNiT2L5SEvOgTDc",{"_type":20,"current":275},"mixture-of-experts","mixture of experts",{"_createdAt":43,"_id":44,"_rev":45,"_type":46,"_updatedAt":43,"slug":278,"title":48},{"current":48},{"_createdAt":43,"_id":50,"_rev":45,"_type":46,"_updatedAt":43,"slug":280,"title":52},{"current":52},"How do mixture-of-experts layers affect transformer models?"]